| ## Generation procedure | |
| The dataset was constructed using documents from [the Pile](https://pile.eleuther.ai/) scored using [LDNOOBW](https://github.com/LDNOOBW/List-of-Dirty-Naughty-Obscene-and-Otherwise-Bad-Words) wordlist (a score is number of curses per character). | |
| The procedure was the following: | |
| 1. The first half of the data are 100k documents randomly sampled from the Pile and assigned scores | |
| 2. The second half are the most cursing document from the Pile, obtained by scoring the whole Pile and choosing 100k documents with highest scores | |
| 3. Then, the dataset was shuffled and a 9:1 train-test split was done | |
| ## Basic stats | |
| The average and median scores are 0.013 and 0.019, respectively. |